Hands-On Data Analysis with Pandas

Stochastic gradient descent classifier

Now that we have the PartialFitPipeline class, the last piece that remains is to select a model capable of online learning. Our logistic regression model performed well—it met the requirements for recall and precision. However, the LogisticRegression class does not support online learning because the method it uses to calculate the coefficients is a closed-form solution. We have the option of using an optimization algorithm, such as gradient descent, to determine the coefficients instead; this will be capable of online learning.

Rather than use a different incremental learner, we can train a new logistic regression model with the SGDClassifier class. It uses stochastic gradient descent (SGD) to optimize ...